Re: New security extension: scripthash

On Sat, Dec 02, 2000 at 10:42:26PM -0000, John Sutton wrote: > > I guess my question was 'how'. SQL databases attempt to do it using a > > user/password mechanism (it's true that with gdbm you won't have that kind > > of protection), but then, that's the mechanism you're trying to > > protect. How does scripthash do it? Maybe I'm just slow, but I didn't quite get it. Apache generates a random secret at startup, right? So the different apps (e.g. mysqlpassd) needs to know this secret to validate the requests, right? So how is the secret shared between the apps and apache? And what prevents a third party application from doing the exact same thing to get the secret? > I have put in the hooks to allow the scripthash user to be the VirtualHost > User rather than, or in addition to, the owner of the script. > > That's the theory! Question is, is it sound? I think yes. (but then again, I didn't even understand it, so who am I to say :-) ). Some detailed issues though: * mysqlpasswd has nothing to do with mysql - maybe a different name would be better? * Could mysqlpassdbm replace the password file instead of changing it? Then it didn't have to stop mysqlpasswd while updating. It would be enough to HUP mysqlpasswd after updating the file - or even better, the deamon can stat the file and discovere it is updated itself. (or maybe even better, mysqlpasswd can compile the dbm file itself from a plaintext file, every time the file is updated - like sendmail does for it's aliases files). * the spec file should put the scripthash files in a seperate rpm from php itself - like the redhat rpm does with all the modules. -- Ragnar Kjørstad Zet.no